Show simple item record

dc.contributor.advisorJaya, I Nengah Surati
dc.contributor.advisorIlham, Qori Pebrial
dc.contributor.authorMayasafitri, Mutmainnah
dc.date.accessioned2026-08-03T14:17:07Z
dc.date.available2026-08-03T14:17:07Z
dc.date.issued2026
dc.identifier.urihttp://repository.ipb.ac.id/handle/123456789/176935
dc.description.abstractPenelitian ini bertujuan mengembangkan algoritma pohon keputusan untuk mendeteksi kesehatan mangrove menggunakan pendekatan non-parametrik. Peubah penginderaan jauh yang digunakan meliputi NDVI, CMRI, EMI, GCI, EVI, BSI, NDMI dan MNDWI yang diturunkan dari citra Sentinel-2A. Sementara itu, peubah sosio-geo-biofisik meliputi jarak dari jalan, pemukiman, sungai, garis pantai, elevasi, kelerengan, dan substrat. Algoritma dibangun menggunakan metode pohon keputusan dengan beberapa kombinasi kriteria pemilihan fitur, yaitu IG, GI, GR, R, dan BF serta optimasi parameter meliputi sampling, pruning, pre-pruning, dan cross validation. Model terbaik diperoleh menggunakan kriteria gini index dengan akurasi model sebesar 96,6%. Hasil uji akurasi klasifikasi menghasilkan overall accuracy sebesar 96,4% dan kappa accuracy sebesar 0,96 yang menunjukkan tingkat kesesuaian sangat kuat antara hasil klasifikasi dan data referensi. Peubah spektral paling berpengaruh dalam memisahkan kesehatan mangrove adalah NDVI diikuti peubah sosio-geo-biofisik berupa substrat. Integrasi indeks spektral dan peubah sosio-geo-biofisik terbukti mampu meningkatkan klasifikasi kesehatan mangrove pada lingkungan pesisir yang kompleks di Kota Batam.
dc.description.abstractThis paper describes a development of machine learning algorithms for detecting mangrove health index by using non-parametric approach. The remotely sensed variables include NDVI, CMRI, EMI, GCI, EVI, BSI, NDMI, and MNDWI indices that derived from Sentinel-2A, while the socio-geo-biophysical data include distance from roads, settlements, rivers, coastline, elevation, slope, and substrate. The algorithm was developed using decision trees with several parameters combination: IG, GI, GR, R, and BF, as well as sampling, pruning, pre-pruning and cross validation. The best model was obtained using the gini index criterion with an accuracy of 96.6%. Classification accuracy assessment produced an overall accuracy of 96.4% and a kappa accuracy of 0.96 indicating a very strong agreement between classification results and reference data. The most influential spectral variable in separating mangrove health classes was NDVI, followed by the socio geo-biophysical variable of substrate. The integration of spectral indices and socio geo-biophysical aspects effectively improved mangrove health classification in the complex coastal environment of Batam City.
dc.description.sponsorship
dc.language.isoid
dc.publisherIPB Universityid
dc.titlePengembangan Algoritma Deteksi Kesehatan Mangrove Berbasis Penginderaan Jauh dan Machine Learning di Kota Batamid
dc.title.alternativeDevelopment of Mangrove Health Detection Algorithm Based on Remote Sensing and Machine Learning in Batam City
dc.typeSkripsi
dc.subject.keywordKesehatan mangroveid
dc.subject.keywordpohon keputusanid
dc.subject.keywordsosio-geo-biofisikid
dc.subtypeUndergraduate Theses


Files in this item

Thumbnail
Thumbnail
Thumbnail

This item appears in the following Collection(s)

Show simple item record